---
license: mit
datasets:
- dvk65/TrashTypes
language:
- en
base_model:
- microsoft/resnet-50
---
This model is trained on a curated dataset of most frequently seen trash items in our college.</br>
## Model Details

- **Backbone**: ResNet50 (ImageNet pre-trained, fine-tuned)
- **Classes**: 13 trash / recycling / compost categories
- **Input size**: 224×224 RGB
- **Loss**: sparse_categorical_crossentropy
- **Optimizer**: Adam

## Dataset

Processed training, validation, and test splits are included in the `*_processed` directories.
Original dataset: [`dvk65/TrashTypes`](https://huggingface.co/dvk65/TrashTypes)

## Usage

```python
from huggingface_hub import hf_hub_download
import tensorflow as tf  # tensorflow version above 2.20.0

REPO_ID = "dvk65/trash-classifier-resnet50"
FILENAME = "trashclassify_13.keras"

model_path = hf_hub_download(
    repo_id=REPO_ID,
    filename=FILENAME,
)

model = tf.keras.models.load_model(model_path)
```
The current target values are:
1. apples
2. bananas
3. bottles
4. cans
5. cardboard
6. cups
7. eggshells
8. mixed leftover food (labeled as generalcompost)
9. wooden coffee stirrers (labeled as mixers)
10. oranges (labeled as peels)
11. platicbags
12. plastic wrappers (labeled as plastics)
13. tissue papers

To help with expanding the dataset, feel free to contribute to: https://huggingface.co/datasets/dvk65/TrashTypes </br>